Deep-learning analysis of 3D microarchitectural remodeling in hypertrophic cardiomyopathy.

Wei, E. Q., Beyer, M., Brown, K. J., Bansbach, A. J., Gorham, J. M., McDonough, B., Chen, H., Khoramjoo, M., Zhang, A., Bishop, B., Ahmad, F., Del Rio, C., Chang, C.-P., Ryba, D. M., Day, S. M., Fatkin, D., Oudit, G. Y., Seidman, C. E., & Seidman, J. G. (2026). Deep-learning analysis of 3D microarchitectural remodeling in hypertrophic cardiomyopathy.. Science (New York, N.Y.), 391(6782), eady6443.

Abstract

Hypertrophic cardiomyopathy (HCM), a genetic heart disease defined by unexplained cardiac wall thickening, is a leading cause of sudden death worldwide. However, the three-dimensional organization of cardiac tissue underlying left ventricular hypertrophy remains poorly understood. We developed CaMVIA-3D, a deep-learning volumetric imaging and analysis pipeline to characterize cardiac microarchitecture. Analysis of tissues from HCM hearts revealed genotype-specific differences in cardiomyocyte volume, morphology, and extracellular volume, with pathogenic variants exhibiting greater concentric cellular hypertrophy and disarray and variant-negative cases showing predominant fibrosis. Longitudinal profiling of a pig HCM model revealed early-onset fibrosis preceding cardiomyocyte hypertrophy. Integrating transcriptomic and morphologic changes, we identified genes associated with cellular and extracellular remodeling. These findings define genotype-specific microstructural differences in HCM, offering insights to improve diagnostics and targeted therapies.

Last updated on 04/01/2026
PubMed